Comparative Analysis of Distinct Fusion Levels in Multimodal Biometrics

Nowadays, Multimodal biometrics has created a substantial interest in the field of identification management due to higher recognition performance. This paper presents a comparative analysis of different fusion levels like feature level, score level and decision level in multimodal biometrics using fingerprint and face. Histogram of Oriented Gradients (HOG) descriptor has been used for fingerprint recognition, Linear Discriminant analysis (LDA) along with Principal component analysis (PCA) for feature reduction and face recognition. These modalities are combined at different fusion levels and the results have shown that biometric fusion at feature level gives superior performance when compared to score level and decision level.